You may be aware of reports suggesting a low prevalence of smoking amongst people hospitalised with COVID-19. Farsalinos and colleagues reviewed data for 5960 patients in China, and the US Centres for Disease Control (CDC) released data for 7162 patients in the USA.
Briefly, smoking prevalence in China is 26.6%, 50.5% in males and 2.1% in females. Farsalinos searched PubMed on 1 April (truly). From 432 studies, they identified 13 that reported smoking status of hospitalized COVID-19 patients. The pooled prevalence of current smoking was 6.5%. In other words it looked as though Chinese people who smoke were less likely to be in hospital with COVID-19.
Likewise in the USA, where adult smoking prevalence in 2018 was 13.7%, the CDC reported on 31 March that only 1.3% of those in hospital with COVID-19 were current smokers.
Both these reports are cross-sectional in design, i.e., they offer a snapshot of the data at a single point in time. Cross-sectional designs are the weakest form of observational study in epidemiology. Taken on their own, these data cannot prove that smoking protects against COVID-19 hospitalisation.
A significant weakness of the reports is that they are not accompanied by a statement of reliability of the data on current smoking. The reliability of hospital inpatient smoking surveys is poorly reported in the literature as a whole. Twenty one years ago a study from Australia found that about two-thirds of smokers could be correctly identified at admission. One can’t assume that such a figure holds true for China or the USA in the context of a pandemic crisis. In the absence of reliability data, “current smoking status”, a primary endpoint for these analyses, has to be viewed as unvalidated.
For example in the case of the Chinese data, male gender in that country is known to be strongly associated with smoking. Chinese COVID-19 hospitalisations by gender reveal a 23% excess for male (3286) versus female (2673). Gender is a robust classifier. It is much more likely to be accurately and thoroughly recorded, than is self-reported smoking status at admission. If smoking were truly protective against hospitalisation, one would expect to see more Chinese women than men admitted with COVID-19. The reverse is true.
Another thing to undermine inference from smoking status to COVID-19 hospitalisation would be any difference in structure of the reported populations from the national average. For example it is possible the studied populations might have lower smoking prevalence in relation to factors such as age, ethnicity, or economic status. Such information is not available in the reports.
In general it is understood that smoking increases the risk of hospitalisation with respiratory disease. For example, in 2019 Han and colleagues found that current smoking increased the odds of influenza hospitalisation by 50% (odds ratio of 1.5, 95% confidence interval 1.3–1.9). This work was based on a pooled analysis from 12 studies, some using the more reliable case-control designs.
Farsalinos et al hypothesise that nicotine might be protective against SARS-CoV-2. There is a prior literature about nicotine’s effects on expression of angiotensin converting enzyme 2 (ACE2) receptors, that the virus uses for cell entry. This is an hypothesis worthy of study. But obviously smoking involves many exposures, of which nicotine is only a part. There could be any number of mechanisms at work.
My view as of 6 April is that it is still not possible to define a strong link between smoking and COVID-19 hospitalisation, in either direction. The virus strikes smokers and non-smokers alike. It is likely to be a different story when it comes to morbidity and survival however. Based on what we know about outcomes for the critically ill who smoke, we have to expect these will be worse.